Cocosplate AI

Create UML activity diagrams in your browser which can actually run. They may access incoming e-mail, wait for client responses and use local or remote AI chatbot endpoints for natural language processing.

Model decisions, receive, send and move E-Mails, perform sentiment analysis on texts and even include custom Python scripts. Cocosplate is a modern, easy to use platform that visualizes your workflows and makes them easier to understand. It provides simple intranet deployment and an affordable license model tailored to medium-sized businesses.

 

3 steps to set up Cocosplate with Docker

By using Docker you can easily setup and maintain an installation. Docker can be downloaded from https://www.docker.com/products/docker-desktop for an installation on Windows or Mac OS X systems. If you are familar with Linux, you can use the Docker installation provided by your favorite distribution.

  1. Get the file cocosplate.zip and extract it
    you can validate the integrity by checking the sha256 sum 4d2f2beac05b4aa59c4578d4ffdce9cbcde608740f06759ba16dbaa18817aed4
  2. Start the system by issuing the following command in the cocosplate directory of the zip-file:
    docker compose up -d
  3. This will download the latest images and start the server.
  4. Open your browser on http://localhost:8081, it will show a login screen.

You can then log in with the username admin and the initial password Login-AI-Workflow-257 and test the system before obtaining a license.

Features

This paragraph shows some of the highlights of Cocosplate AI available to date. Since the activity diagrams can be customized your possibilities of creating workflows are almost endless.

Use Large Language Models in automated AI processing and sentiment detection for customer care

The AI workbench allows to immediately evaluate LLM prompts and create a processing node from it in your activity diagram.
Language model prompt workbenchThe system currently supports the OpenAI API (subscription required) and additionally is able to operate on a local or intranet-accessible language model such as llama2 with GPT4all on which you have full control. This allows you to create a setup compliant to the European GDPR.
The sentiment detection module has many use-cases. One is the detection of negative customer input to determine that a request should not be answered in an automated way.

Engine support for asynchronous e-mail answer clicks

The system can be configured to receive clicks and operate based on the user-requested input.
In this example the user may accept or decline a request sent via e-mail by clicking links in the message. A timeout event occurs after a certain timespan which may consider the request obsolete or in this case as declined, too.

 

A template editor with an intuitive interface allows to edit and maintain custom texts for for bulk processing. It supports quick links to insert available actions and variables derived from the diagram.
The template may also contain macros to determine which text snippets to use. The example shows a gender-based salutation. All data processed within the workflow can be accessed and inserted into the text.

 

 

Python scripting

The system supports python scripts that are able to use the data in the workflow. The created output may be used in consecutive steps in the diagram. That way you are able to extend any workflow to your needs and even determine which exit to take.